Module 6: Dimensionality Reduction
  Lecture 33: Embedding
 

                                            

 

 

Embedding Quadratic Form distance
  • QFDs and other statistical distances do not allow indexes
 
  • Aim is to design a low distortion low-dimensional embedding
  • Matrix in QFD allows Cholesky decomposition since is symmetric and positive definite
                               
                                                
                                                   
  • Therefore, embedding is
  • Allows further dimensionality reduction as well